Social Media Aided Sentiment Analysis for Stock Prediction

Abstract

In the stock market research, stock prediction is a challenging task due to its dynamic characteristic very similar to wealth of a nation and opinion about a stock. It is very difficult for the investor to buy or sell their stock because of noisy, chaotic properties of the stock data. Stock prediction mostly performed depends up on the numerical data obtained with technical measures or text data provided by the data sources as sentiments. A change in the fundamental measures obtained from exchange rate , gold price and crude oil price also determines the stock value. User generated contents of sentiments available in various social media like Twitter and News sites also play an important role for deciding the price of the stock. Most of the existing work deals any one of the measures technical measures or fundamental measures or sentiment measures for predicting the price of the stock. Hence, the proposed method employs combined measures derived from technical, fundamental and sentiments. Twitter and Money Control act as a data source for providing opinion data to predict the stock price. Results of the proposed system compared with the others by using various measures such as accuracy, sensitivity, specificity, Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). It is found that the proposed methods for stock prediction outperform the existing techniques.